Table 2.

Artificial neural network predictions for caffeine, theobromine and paraxanthine. ANNs were constructed with either a single output node (one model for each of the analytes) or using three output nodes in a single ANN

	Caffeine	Theobromine	Paraxanthine
Single output node in a 1024-20-1 topology
R 2	0.9999	0.9939	0.9884
Q 2	0.7067	0.8958	0.8953
RMSEP (mol/dm3)	1.6832 × 10−6	8.8402 × 10−7	9.6140 × 10−7
Three output nodes in a 1024-20-3 topology
R 2	0.9999	0.9999	0.9999
Q 2	0.8093	0.8115	0.8785
RMSEP (mol/dm3)	1.3222 × 10−6	1.3145 × 10−6	1.0553 × 10−6